VoiceAI Engineer (LLM, NLP)

Trusting Social is an AI Fintech pioneer that's revolutionizing credit access in emerging markets. Our mission is "Advancing AI to Meet the Financial Needs of Everyday Consumers with Empathy." We've assessed over 1 billion consumers across four countries, and we're on a mission to provide 100 million credit lines using the power of AI and Big Data.How You'll Make an ImpactAs a Voice AI Engineer, you'll go beyond building models - you'll architect our "Digital Humans." You'll design the orchestration layer that enables our AI agents to listen, reason, and respond in real-time, seamlessly connecting rigid banking systems with the natural flow of human conversation in financial advisory scenarios.What You'll DoYou'll spearhead the development of our real-time omnichannel communication platform, focusing on creating voice agents that feel truly human: low-latency, intuitive, and capable of handling nuanced financial discussions with empathy and precision.End-to-end pipeline orchestration: Design and optimize the complete conversational flow: voice activity detection (VAD) → automatic speech recognition (ASR) → large language model (LLM) reasoning → text-to-speech (TTS).Latency optimization: Target sub-1200ms end-to-end perceived latency using techniques like stream-to-stream processing.Advanced speech understanding: Fine-tune a speech foundation model with multiple speech understanding capabilities, including automatic speech recognition (ASR), spoken language identification (LID), speech emotion recognition (SER), and audio event detection (AED)Advanced speech generation: TTS models for controllably expressive and emotional speech generation: natural prosody, emotional tone, and adaptation to local Southeast Asian dialects and accents. Develop voice cloning for consistent, brand-aligned agents that convey warmth during sensitive financial topics.Turn-taking & interruptibility: Build robust logic for handling interruptions (barge-in), background noise, filler words, and natural pauses, ensuring smooth, human-like dialogue.Agent memory: Embed memory into the voice pipeline for real-time, accurate delivery of personalized communication across channels, without disrupting conversation flow.System reliability & evaluation: Develop agent versioning, conduct A/B testing, and track key internal voice agent / biz-related metrics for user engagement to continuously improve interactions.